// This file is part of BOINC.
// http://boinc.berkeley.edu
// Copyright (C) 2011 University of California
//
// BOINC is free software; you can redistribute it and/or modify it
// under the terms of the GNU Lesser General Public License
// as published by the Free Software Foundation,
// either version 3 of the License, or (at your option) any later version.
//
// BOINC is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
// See the GNU Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public License
// along with BOINC. If not, see .
// process_input_template():
// fill in the workunit's XML document (wu.xml_doc)
// by scanning the input template, macro-substituting the input files,
// and putting in the command line element and additional XML
#include
#include
#include
#include
#include
#include "filesys.h"
#include "md5_file.h"
#include "str_replace.h"
#include "sched_config.h"
#include "sched_util.h"
#include "process_input_template.h"
using std::string;
using std::vector;
// look for file named FILENAME.md5 containing md5sum and length.
// If found, and newer mod time than file,
// read md5 sum and file length from it, and return true.
// Else return false
//
// See checkin notes Dec 30 2004
//
static bool got_md5_info(
const char *path,
char *md5data,
double *nbytes
) {
char md5name[512];
struct stat md5stat, filestat;
char endline='\0';
sprintf(md5name, "%s.md5", path);
// get mod times for file
//
if (stat(path, &filestat)) {
return false;
}
// get mod time for md5 cache file
//
// coverity[toctou]
if (stat(md5name, &md5stat)) {
return false;
}
// if cached md5 newer, then open it
#ifndef _USING_FCGI_
FILE *fp=fopen(md5name, "r");
#else
FCGI_FILE *fp=FCGI::fopen(md5name, "r");
#endif
if (!fp) {
return false;
}
// read two quantities: md5 sum and length.
// If we can't read these, or there is MORE stuff in the file,
// it's not an md5 cache file
//
int n = fscanf(fp, "%s %lf%c", md5data, nbytes, &endline);
int c = fgetc(fp);
fclose(fp);
if ((n != 3) || (endline !='\n') || (c != EOF)) {
fprintf(stderr, "bad MD5 cache file %s; remove it and stage file again\n", md5name);
return false;
}
// if this is one of our cached md5 files, but it's OLDER than the
// data file which it supposedly corresponds to, delete it.
//
if (md5stat.st_mtime element for workunit XML doc,
// based on the input template and list of variable files
//
// Inputs:
// xp: parser for input template
// var_infiles: list of files descs passed to create_work (i.e. var files)
//
// Outputs:
// infiles: vector (appended to) of all input files
// out:
//
// Actions:
// as an input file for create_work.
// out: append the element for the WU XML doc
// If not a constant file:
// increment nfiles_parsed
//
static int process_file_info(
XML_PARSER& xp, SCHED_CONFIG& config_loc,
vector &var_infiles,
string& out,
vector &infiles,
int& n_var_infiles
) {
vector urls;
bool gzip = false;
int retval, itemp;
double nbytes, nbytesdef = -1, gzipped_nbytes;
string md5str, urlstr, tmpstr;
char physical_name[256];
char buf[BLOB_SIZE], path[MAXPATHLEN], top_download_path[MAXPATHLEN];
char gzip_path[MAXPATHLEN];
char md5[33], url[256], gzipped_url[256], buf2[256];
INFILE_DESC var_infile, infile;
strcpy(physical_name, "");
out += "\n";
while (!xp.get_tag()) {
if (xp.parse_bool("gzip", gzip)) {
continue;
} else if (xp.parse_str("physical_name", physical_name, sizeof(physical_name))) {
continue;
} else if (xp.parse_string("url", urlstr)) {
urls.push_back(urlstr);
continue;
} else if (xp.parse_int("number", itemp)) {
// ignore
continue;
} else if (xp.parse_string("md5_cksum", md5str)) {
continue;
} else if (xp.parse_double("nbytes", nbytesdef)) {
continue;
} else if (xp.parse_double("gzipped_nbytes", gzipped_nbytes)) {
continue;
} else if (xp.match_tag("/file_info")) {
// there are four cases:
// - normal file
// - constant file
// is given
// - remote file, specified as create_work() arg
// URL, size etc. are given in INFILE_DESC
// - remote file, specified in template
// URL, size etc. are given in template
//
if (!strlen(physical_name)) {
if (n_var_infiles >= (int)var_infiles.size()) {
fprintf(stderr,
"Too few var input files given; need at least %d\n",
n_var_infiles+1
);
return ERR_XML_PARSE;
}
var_infile = var_infiles[n_var_infiles];
}
if (strlen(physical_name)) {
// constant file case
//
dir_hier_path(
physical_name, config_loc.download_dir,
config_loc.uldl_dir_fanout, path, true
);
if (!got_md5_info(path, md5, &nbytes)) {
fprintf(stderr, "missing MD5 info file for %s\n",
physical_name
);
return ERR_FILE_MISSING;
}
dir_hier_url(
physical_name, config_loc.download_url,
config_loc.uldl_dir_fanout, url
);
sprintf(buf,
" %s\n"
" %s\n"
" %s\n"
" %.0f\n",
physical_name,
url,
md5,
nbytes
);
strcpy(infile.name, physical_name);
strcpy(infile.md5, md5);
infile.nbytes = nbytes;
} else if (nbytesdef > 0) {
// remote file specified in template
//
if (md5str == "" || urlstr == "") {
fprintf(stderr, "All file properties must be defined "
"if at least one is defined (url, md5_cksum, nbytes)\n"
);
return ERR_XML_PARSE;
}
if (gzip && !gzipped_nbytes) {
fprintf(stderr, "Must specify gzipped_nbytes\n");
return ERR_XML_PARSE;
}
urlstr = "";
for (unsigned int i=0; i" + urls.at(i) + string(var_infile.name) + "\n";
if (gzip) {
urlstr += " " + urls.at(i) + string(var_infile.name) + ".gz\n";
}
}
sprintf(buf,
" %s\n"
"%s"
" %s\n"
" %.0f\n",
var_infile.name,
urlstr.c_str(),
md5str.c_str(),
nbytesdef
);
if (gzip) {
sprintf(buf2, " %.0f\n",
gzipped_nbytes
);
strcat(buf, buf2);
}
strcpy(infile.name, var_infile.name);
strcpy(infile.md5, md5str.c_str());
infile.nbytes = nbytesdef;
} else if (var_infile.is_remote) {
// remote file specified in create_work() arg
//
sprintf(buf2, "jf_%s", var_infile.md5);
sprintf(buf,
" %s\n"
" %s\n"
" %s\n"
" %.0f\n",
buf2,
var_infile.url,
var_infile.md5,
var_infile.nbytes
);
strcpy(infile.name, buf2);
strcpy(infile.md5, var_infile.md5);
infile.nbytes = var_infile.nbytes;
} else {
// normal case. we need to find file size and MD5;
// stage the file if needed
//
dir_hier_path(
var_infile.name, config_loc.download_dir,
config_loc.uldl_dir_fanout, path, true
);
// if file isn't found in hierarchy,
// look for it at top level and copy
//
if (!boinc_file_exists(path)) {
sprintf(top_download_path,
"%s/%s",config_loc.download_dir, var_infile.name
);
boinc_copy(top_download_path, path);
printf("copy %s to %s\n", top_download_path, path);
}
if (!config_loc.cache_md5_info || !got_md5_info(path, md5, &nbytes)) {
retval = md5_file(path, md5, nbytes);
if (retval) {
fprintf(stderr,
"process_input_template: md5_file %s\n",
boincerror(retval)
);
return retval;
} else if (config_loc.cache_md5_info) {
write_md5_info(path, md5, nbytes);
}
}
dir_hier_url(
var_infile.name, config_loc.download_url,
config_loc.uldl_dir_fanout, url
);
if (gzip) {
sprintf(gzip_path, "%s.gz", path);
retval = file_size(gzip_path, gzipped_nbytes);
if (retval) {
fprintf(stderr,
"process_input_template: missing gzip file %s\n",
gzip_path
);
return ERR_FILE_MISSING;
}
sprintf(gzipped_url,
" %s.gz\n"
" %.0f\n",
url, gzipped_nbytes
);
} else {
strcpy(gzipped_url, "");
}
sprintf(buf,
" %s\n"
" %s\n"
"%s"
" %s\n"
" %.0f\n",
var_infile.name,
url,
gzipped_url,
md5,
nbytes
);
strcpy(infile.name, var_infile.name);
strcpy(infile.md5, md5);
infile.nbytes = nbytes;
}
infiles.push_back(infile);
if (!strlen(physical_name)) {
n_var_infiles++;
}
out += buf;
out += "\n";
break;
} else {
// copy any other elements from input template to XML doc
//
retval = xp.copy_element(tmpstr);
if (retval) return retval;
out += tmpstr;
out += "\n";
}
}
return 0;
}
// parse the elements, which includes file refs and job params
//
static int process_workunit(
XML_PARSER& xp, WORKUNIT& wu, string& out,
vector &infiles,
const char* command_line,
const char* additional_xml
) {
char buf[256], open_name[256];
string tmpstr, cmdline;
int retval, n_file_refs=0, itemp;
out += "\n";
if (command_line) {
//fprintf(stderr, "appending command line: %s\n", command_line);
out += "\n";
out += command_line;
out += "\n\n";
}
while (!xp.get_tag()) {
if (xp.match_tag("/workunit")) {
if (additional_xml && strlen(additional_xml)) {
out += additional_xml;
out += "\n";
}
out += "";
break;
} else if (xp.match_tag("file_ref")) {
out += "\n";
bool found_open_name = false, found_copy_file = false;
if (n_file_refs >= (int)infiles.size()) {
fprintf(stderr, "too many s\n");
return ERR_XML_PARSE;
}
INFILE_DESC& id = infiles[n_file_refs];
sprintf(buf, " %s\n", id.name);
out += buf;
while (!xp.get_tag()) {
if (xp.parse_str("open_name", open_name, sizeof(open_name))) {
sprintf(buf, " %s\n", open_name);
out += buf;
found_open_name = true;
continue;
} else if (xp.parse_int("file_number", itemp)) {
} else if (xp.match_tag("copy_file/")) {
out += " \n";
found_copy_file = true;
continue;
} else if (xp.match_tag("/file_ref")) {
if (!found_open_name && !found_copy_file) {
fprintf(stderr, "No open name found and copy_file not specified\n");
return ERR_XML_PARSE;
} else if (!found_open_name && found_copy_file) {
sprintf(buf, " %s\n", id.name);
out += buf;
}
out += "\n";
break;
} else if (xp.parse_string("file_name", tmpstr)) {
fprintf(stderr, " ignored in element.\n");
continue;
} else {
retval = xp.copy_element(tmpstr);
if (retval) return retval;
out += tmpstr;
out += "\n";
}
}
n_file_refs++;
} else if (xp.parse_string("command_line", cmdline)) {
if (command_line) {
fprintf(stderr, "Can't specify command line twice\n");
return ERR_XML_PARSE;
}
out += "\n";
out += cmdline;
out += "\n\n";
} else if (xp.parse_double("rsc_fpops_est", wu.rsc_fpops_est)) {
continue;
} else if (xp.parse_double("rsc_fpops_bound", wu.rsc_fpops_bound)) {
continue;
} else if (xp.parse_double("rsc_memory_bound", wu.rsc_memory_bound)) {
continue;
} else if (xp.parse_double("rsc_bandwidth_bound", wu.rsc_bandwidth_bound)) {
continue;
} else if (xp.parse_double("rsc_disk_bound", wu.rsc_disk_bound)) {
continue;
} else if (xp.parse_int("batch", wu.batch)) {
continue;
} else if (xp.parse_int("delay_bound", wu.delay_bound)){
continue;
} else if (xp.parse_int("min_quorum", wu.min_quorum)) {
continue;
} else if (xp.parse_int("target_nresults", wu.target_nresults)) {
continue;
} else if (xp.parse_int("max_error_results", wu.max_error_results)) {
continue;
} else if (xp.parse_int("max_total_results", wu.max_total_results)) {
continue;
} else if (xp.parse_int("max_success_results", wu.max_success_results)) {
continue;
} else if (xp.parse_int("size_class", wu.size_class)) {
continue;
} else {
retval = xp.copy_element(tmpstr);
if (retval) return retval;
out += tmpstr;
out += "\n";
}
}
if (n_file_refs != (int)infiles.size()) {
fprintf(stderr, "#file refs != #file infos\n");
return ERR_XML_PARSE;
}
return 0;
}
// fill in the workunit's XML document (wu.xml_doc)
// by scanning the input template, macro-substituting the input files,
// and putting in the command line element and additional XML
//
int process_input_template(
WORKUNIT& wu,
char* tmplate,
vector &var_infiles,
// files passed as args to create_work
SCHED_CONFIG& config_loc,
const char* command_line,
const char* additional_xml
) {
string out;
int retval;
bool found_workunit=false;
vector infiles;
// this gets filled in as we scan s
int n_var_infiles=0;
// number of non-constant infiles scanned
// this is an index into var_infiles
// "out" is the XML we're creating
//
out = "";
MIOFILE mf;
XML_PARSER xp(&mf);
mf.init_buf_read(tmplate);
while (!xp.get_tag()) {
if (!xp.is_tag) continue;
if (xp.match_tag("input_template")) continue;
if (xp.match_tag("/input_template")) continue;
if (xp.match_tag("file_info")) {
retval = process_file_info(
xp, config_loc, var_infiles,
out, infiles, n_var_infiles
);
if (retval) return retval;
} else if (xp.match_tag("workunit")) {
found_workunit = true;
retval = process_workunit(
xp, wu, out, infiles, command_line, additional_xml
);
if (retval) return retval;
}
}
if (!found_workunit) {
fprintf(stderr, "process_input_template: bad WU template - no \n");
return ERR_XML_PARSE;
}
if (n_var_infiles != (int)var_infiles.size()) {
fprintf(stderr,
"process_input_template: %d input files given, but template has %d\n",
(int)var_infiles.size(), n_var_infiles
);
return ERR_XML_PARSE;
}
if (out.size() > sizeof(wu.xml_doc)-1) {
fprintf(stderr,
"create_work: WU XML field is too long (%d bytes; max is %d)\n",
(int)out.size(), (int)sizeof(wu.xml_doc)-1
);
return ERR_BUFFER_OVERFLOW;
}
//fprintf(stderr, "copying to xml_doc: %s\n", out.c_str());
safe_strcpy(wu.xml_doc, out.c_str());
return 0;
}
#ifdef TEST
SCHED_CONFIG config_loc;
int main(int, char**) {
WORKUNIT wu;
vector var_infiles;
char* tmplate;
const char* command_line = "sample command line";
const char* additional_xml;
config_loc.parse_file("");
read_file_malloc("template_in", tmplate);
INFILE_DESC id;
#if 1
strcpy(id.name, "input");
var_infiles.push_back(id);
#endif
#if 0
strcpy(id.name, "input");
id.is_remote = true;
strcpy(id.url, "http://blah.foo");
strcpy(id.md5, "sfslkjsdlfkj");
id.nbytes = 234;
var_infiles.push_back(id);
#endif
int retval = process_input_template(
wu,
tmplate,
var_infiles,
config_loc,
command_line,
additional_xml
);
if (retval) {
printf("process_input_template returned %d\n", retval);
} else {
printf("xml_doc:\n%s\n", wu.xml_doc);
}
}
#endif